High ISO Performance and Fast Lenses

[Whiskeyman]

A coworker and I had a discussion that we couldn't decide the answer to over lunch today. Can you folks chime in with your thoughts.

It started when I made the statement that cameras with high ISO performance were getting extremely good, and less expensive, and may soon get to the point that really fast glass was going to become unnecessary. We each found arguments against and for the statement, by the way.

Do yo think that we're going to reach the point where a f/2.8 (or faster) telephoto lens is going to become obsolete due to cost and the high ISO performance of cameras?

WM

 

[WayneF]

Do yo think that we're going to reach the point where a f/2.8 (or faster) telephoto lens is going to become obsolete due to cost and the high ISO performance of cameras?

Not if performance matters more than price. It is not only about exposure.

A f/2.8 lens will do f/2.8 (depth of field, etc). A f/4 lens will not.

But most important (to me) is that a f/2.8 lens stopped down to f/4 is sharper and performs better at f/4 than a f/4 lens wide open (for bounce flash, etc). Price buys a lot.

 

[kevy73]

Speed isn't as important to me as narrow depth of field... I don't use 1.4 because I need the shutter speed, I use it because I want that sensational depth of field...

 

[Moab Man]

Leering into my crystal ball... I can see one day where ISO noise is really not an issue. Better faster cleaner sensors are always pushing the envelope. I can see a point where the aperture becomes less and less about light being allowed in, and while not exclusively about depth of field, but heavily disproportionate towards depth of field.

The crystal ball goes hazy. That is all for today unless you want to hand over another $20?

 

[J-see]

Do yo think that we're going to reach the point where a f/2.8 (or faster) telephoto lens is going to become obsolete due to cost and the high ISO performance of cameras?

WM

I think it's quite the opposite. The next (currently released) generation are backside illuminated sensors which will perform better the smaller their sensor pixels. From what I read their best performance would be around 1 micron. The downside is that they'll become more diffraction limited the more that pixel size goes down which will demand lenses that go wider open if you want to have some options left.

Besides that, for similar "action" lenses, f/2.8 will always outperform f/4 when it comes to focus speed. That matters a lot for specific types of photography.

 

[gqtuazon]

A coworker and I had a discussion that we couldn't decide the answer to over lunch today. Can you folks chime in with your thoughts.

Do yo think that we're going to reach the point where a f/2.8 (or faster) telephoto lens is going to become obsolete due to cost and the high ISO performance of cameras?

WM

Not if performance matters more than price. It is not only about exposure.

A f/2.8 lens will do f/2.8 (depth of field, etc). A f/4 lens will not.

But most important (to me) is that a f/2.8 lens stopped down to f/4 is sharper and performs better at f/4 than a f/4 lens wide open (for bounce flash, etc). Price buys a lot.

Totally agree and echo Wayne's comments.

I am not a fan of the variable aperture lenses. My preference is to keep the ISO as low as possible to optimize Dynamic Range and have a much cleaner image especially if you tend to crop a lot.

 

[Felisek]

There is a hard physical limit to which you can improve sensor sensitivity. It is counting noise. When you have only a handful of photons arriving per pixel (and as J-see mentioned, pixels are getting smaller), there will be Poisson noise in the final image, whatever smart technological tricks you apply it. This is basic physics (or rather statistics) and you cannot beat it.

 

[J-see]

There is a hard physical limit to which you can improve sensor sensitivity. It is counting noise. When you have only a handful of photons arriving per pixel (and as J-see mentioned, pixels are getting smaller), there will be Poisson noise in the final image, whatever smart technological tricks you apply it. This is basic physics (or rather statistics) and you cannot beat it.

That's indeed an issue in which the ISO performance of a cam can never outperform the light collection of a lens. Noise = the signal collected by a pixel and it is impossible to distinguish between both even when we talk about them as if they're both separate. The best we can expect of ISO performance is that it does not add more noise and maintains the SNR identical as the signal collected. A wider aperture will always collect more light (at the same shutter speed) and in doing that, improve the SNR and thus deliver a better signal.

It's impossible to get around that.

 

[WayneF]

The downside is that they'll become more diffraction limited the more that pixel size goes down which will demand lenses that go wider open if you want to have some options left.

FWIW, I personally think this pixel size part is a myth about diffraction, only a confusion factor invented and promoted by techie wannabes, proud they learned to compute things (meaningful or not), instead of being promoted by lens or sensor designers who count (and actually know). My opinion, but Nikon, Canon and Sony don't say things like that either.

Diffraction does of course increase with stopped down aperture, but pixel size is not a factor of diffraction or of diffraction effects. Smaller pixels may mean more noise, but smaller pixels are always simply greater resolution (for reproducing the resolution of the lens image), which aspect is always a good thing, at least until up to a point well beyond the lens resolution (Nyquist, etc).

Somehow the techies manage to even assume the airy disk must somehow be perfectly centered on the pixel area (because they can calculate that). But even if it magically in fact were, greater resolution to show what is actually present there is always a good thing (better than being unable to resolve the detail that is there).

Certainly there are times that f/22 or f/32, or even more does greatly improve the image (some images, not all images), even though diffraction obviously increases. More so in telephoto lenses of course, not so much in wide angles, since diffraction is about aperture size, not about pixel size. Telephoto lenses normally do provide f/32 because it certainly can help some cases. Wide angle lens don't, f/32 won't help them much. Diffraction is just a tradeoff with depth of field, which are tools, and there are cases when either one is most important. Diffraction is about aperture size, not pixel size. Film had grain too, small particles comparable to pixels, but the subject never was any issue about diffraction until internet techies. The issue was always about resolving the detail that was there. We ought to question anything we read on the internet. We ought to be able to SEE the things we claim to believe

I put some tries to show this at Diffraction limited images? Really?

It compares a 12 mp DX D300 with a 36 mp FX D800. It easily shows stopping down more can sometimes help the image, and that greater enlargement of smaller sensor size certainly is also a factor, but it is unable to show pixel size has any effect at all. Diffraction is about aperture, and smaller pixel size is simply greater resolution, always a good thing (except for noise). .

 

[J-see]

FWIW, I personally think this pixel size part is a myth about diffraction, only a confusion factor invented and promoted by techie wannabes, proud they learned to compute things (meaningful or not), instead of being promoted by lens or sensor designers who count (and actually know). My opinion, but Nikon, Canon and Sony don't say things like that either.

When the aperture is made smaller by 'stopping down' the lens past a specific aperture diffraction becomes noticable, resulting in a softening of the image and a reduction in resolution. The aperture at which diffraction occurs will vary depending on the combination of lens and camera used. The effects of diffraction are partly influenced by the size of the pixels on the camera’s image sensor.

https://support.nikonusa.com/app/answers/detail/a_id/18344/~/what-is-diffraction?

So much for that myth huh?

 

[WayneF]

https://support.nikonusa.com/app/answers/detail/a_id/18344/~/what-is-diffraction?

So much for that myth huh?

No, so much for the support writers, we know they are capable of saying anything.

If there is any actual evidence, it needs to be shown to us, seeing is believing. IMO, all the actual evidence is contrary.

 

[J-see]

No, so much for the support writers, we know they are capable of saying anything.

If there is any actual evidence, it needs to be shown to us, seeing is believing. IMO, all the actual evidence is contrary.

Any actual evidence? Maybe use google and you'll find out the whole world is aware of diffraction limitation in imagining systems.

I'm kind and will do the work for you:

Implications for digital photography[edit]

In a digital camera, diffraction effects interact with the effects of the regular pixel grid. The combined effect of the different parts of an optical system is determined by the convolution of the point spread functions (PSF). The point spread function of a diffraction limited lens is simply the Airy disc. The point spread function of the camera, otherwise called the instrument response function (IRF) can be approximated by a rectangle function, with a width equivalent to the pixel pitch. A more complete derivation of the modulation transfer function (derived from the PSF) of image sensors is given by Fliegel.[SUP][3][/SUP] Whatever the exact instrument response function we may note that it is largely independent of the f-number of the lens. Thus at different f-numbers a camera may operate in three different regimes, as follows:

1. in the case where the spread of the IRF is small with respect to the spread of the diffraction PSF, in which case the system may be said to be essentially diffraction limited (so long as the lens itself is diffraction limited).
2. in the case where the spread of the diffraction PSF is small with respect to the IRF, in which case the system is instrument limited.
3. in the case where the spread of the PSF and IRF are of the same order of magnitude, in which case both impact the available resolution of the system.

The spread of the diffraction-limited PSF is approximated by the diameter of the first null of the Airy disk,

where λ is the wavelength of the light and N is the f-number of the imaging optics. For f/8 and green (0.5 μm wavelength) light, d = 9.76 μm. This is of the same order of magnitude as the pixel size for the majority of commercially available 'full frame' (43mm sensor diagonal) cameras and so these will operate in regime 3 for f-numbers around 8 (few lenses are close to diffraction limited at f-numbers smaller than 8). Cameras with smaller sensors will tend to have smaller pixels, but their lenses will be designed for use at smaller f-numbers and it is likely that they will also operate in regime 3 for those f-numbers for which their lenses are diffraction limited.

https://en.wikipedia.org/wiki/Diffraction-limited_system


If you are convinced Wikipedia also is capable of "writing anything", tell me and I'll organize a fundraiser to see if I can get an expedition going to Mount Sinai to check if some divine being did cast it in stone.

 

[WayneF]

Any actual evidence? Maybe use google and you'll find out the whole world is aware of diffraction limitation in imagining systems.

I'm kind and will do the work for you:



https://en.wikipedia.org/wiki/Diffraction-limited_system


If you are convinced Wikipedia also is capable of "writing anything", tell me and I'll organize a fundraiser to see if I can get an expedition going to Mount Sinai to check if some divine being did cast it in stone.

LOL. Start your fundraiser then. I am a big fan of Wikipedia, often my first easy source, one of the prime beauties of the internet, but I am very careful about believing all I read, because certainly any of the unidentified writers is capable of writing anything there. And of course they do. It often gets corrected, maybe in months or years, maybe never. Sometimes correct stuff gets corrupted by others. If you want to learn something, on most topics of any difficulty, you should also read under the TALK tab at Wikipedia.

Any and all of the evidence on the internet about pixel limited diffraction is merely carefully hand-prepared graphics designed intentionally to show the point they espouse. CambridgeInColor is a major source of this, but this guy also believes gamma is done for the human eye instead of for CRT displays. (outright laughable) Just pretty graphics, but he just read something somewhere.

Regarding pixel limited diffraction, there is Zero actual evidence of results mattering, shown or otherwise. I think it does not matter and cannot be shown, because, well, you know why. Any actual Airy disk will normally be straddling pixel boundaries (mostly, any pixel size, any aperture size, if not in the one pixel we examine, certainly in all the other megapixels). An Airy disk is only from a point source anyway. Other than stars, we don't photograph many point sources.

But the greater resolution of smaller pixels certainly will always help to reproduce whatever detail is actually there. The job of small pixels is to show more detail. Large pixels without sufficient resolution to show it does NOT mean it is not there, it fails to show many things.

Greater resolution is always a good thing (other than noise).
Greater resolution should show the diffraction better, but that is the detail that is there. The pixels did not cause it, they merely better show what is there. It was already there. Smaller pixels show all detail better. They don't cause the detail, they merely show it. But at greatly reduced normal viewing sizes, smaller original pixels will not show more then.

People worry about the better sensors outresolving their inexpensive lenses. But the fault is the lens, not the sensor. Show the faults, or show less detail so not even the faults can show... which is best performance? Same principle with diffraction. The pixels only show it, they do not cause it. And smaller pixels are greater resolution, more capable to correctly show ALL detail there.

Compact cameras are severely diffraction limited past say f/4. Maybe f/5.6 because we never print them large. This is because of their tiny focal length reducing aperture size so much, not because of tiny pixels. We have had compact sensors sizes starting from 0.3 mp to at least 16 mp now, and the answer is always the same apertures. And diffraction works in the last decade the same way it has always worked for 100+ years.

The internet is different. In the old days, books or encyclopedias or even magazines had signed sources, who were accountable. The internet, wonderful as it is, is NOT accountable for anything. Everything has to be evaluated, and then, still keep an open mind.

Think about things more. Wonder how it can possibly be? Or at least show some actual pictorial evidence of results. If it matters, show it matters. Show what you claim to believe. We ought to be able to see what we claim matters.

-------------------------------------------------
​

MY ONLY POINT about any of this, is that the dumb articles about pixel limited diffraction make many users shy away from ever using an aperture smaller than about f/11. They heard it was bad.

But the obvious truth is, there are MANY situations when the greater depth of field of f/22 or f/32 or even more at times, can make a tremendously better image than the f/11 image where the foreground and background are too fuzzy to recognize. Which is the reason the lens provides those apertures. It is a trade off, sometimes either side can win. But you really ought to try it sometimes, when appropriate.

J-see, You meaning any reader, not meaning only you.


See Diffraction limited images? Really? (and next page too).

 

[J-see]

Evidently the sensor pixels are at the receiving end of diffraction since diffraction is a property of waves. The smaller the aperture, the more those waves spreads. But the catch is that the smaller those pixels at the receiving end, the faster those waves occupy too much surface for the cam to be able to distinguish between the one and the other pixel.

If you heavily down-sample a shot it matters less since perceived sharpness increases while down-sampling but the moment you start cropping it makes all the difference.

Most FX are diffraction limited around f/22. After that there is no gain. But even long before it might become noticeable depending the subject.


If you want to do the test; take a page out of a book and shoot the text at 1:1 using f/4 up to f/40 and then show me the 100% crops.

 

[WayneF]

Evidently the sensor pixels are at the receiving end of diffraction since diffraction is a property of waves. The smaller the aperture, the more those waves spreads. But the catch is that the smaller those pixels at the receiving end, the faster those waves occupy too much surface for the cam to be able to distinguish between the one and the other pixel.

If you heavily down-sample a shot it matters less since perceived sharpness increases while down-sampling but the moment you start cropping it makes all the difference.

Hardly worth the effort if you don't bother to read it. What about some of these points?

1. the diffraction is already there, regardless if the lesser resolution of larger pixels can show it or not. Loss of detail is loss of detail.
2. the 20 years of compact camera sensor resolutions (pixel size) did not make a dent in the diffraction limited useful apertures we can use.
3. we don't even photograph point sources or Airy disks.

Certainly diffraction exists, no one is arguing otherwise. Certainly it is worse at f/40 than f/8, no one is arguing otherwise.

But we do not necessarily drop dead there. It depends, sometimes the additional depth of field helps more than the greater diffraction hurts. This is quite obvious, you should try it. We should think about things, and try some things, and go with what works. Seeing is believing.

Most FX are diffraction limited around f/22. After that there is no gain. But even long before it might become noticeable depending the subject.


If you want to do the test; take a page out of a book and shoot the text at 1:1 using f/4 up to f/40 and then show me the 100% crops.

Simply incorrect, it is the worst possible advice, at least from a photographers viewpoint of wanting to produce better photographs (it does depend on the situation). Really, you actually don't understand that it is also about depth of field?
If f/32 is not in your arsenal, you are going out unarmed (at times).
​

If you want to see my f/40, see Diffraction limited images? Really?

Bottom of page is f/40. There is also some f/32 there and next page. FX and DX, and a few lenses.


I can easily see diffraction. It is a bad thing, but I can also see that sometimes the greater depth of field helps, a lot, helps more than diffraction hurts.

I can easily see having to enlarge a smaller sensor more is less good than less enlargement of larger sensor (but the views are different).

But I can find no evidence of pixel limited diffraction. The diffraction is already there, and the pixel resolution simply shows the details there.

 

[Woodyg3]

As to the original question, I think fast glass will always be desired for depth of field limitation and for indoor sports and similar low light situations. I say always, but we know that current technology will look ancient in 30 years. Just as digital took the place of film, another technology will come along and replace digital. The current limitations of digital sensors may seem just as prehistoric as the grainy black and white photos I used to produce by push processing Tri-X Pan to 1600 ASA.

As to the whole worry about diffraction, let's remember that we tend not to print 100% crops and hang them on our walls.

As a practical example, photographers have been using small apertures for landscapes since the beginning of photography in order to get a wide depth of field. It works fine because nobody stands 6 inches away from a print and looks at it with magnifying glass. Only photo geeks like me look at 100% crops to judge lens and camera performance.

 

[WayneF]

I am glad to see a thinking reasonable view.

 

[J-see]

I can easily see diffraction. It is a bad thing, but I can also see that sometimes the greater depth of field helps, a lot, helps more than diffraction hurts.

Hola; let's pause for a second.

Who's talking about DoF?

 

[WayneF]

Hola; let's pause for a second.

Who's talking about DoF?

Aw come on J-see, I doubt you really are that obtuse. The issue is about conditions for creating a better image, and diffraction generally detracts (true of the step from f/8 to f/11 too). However DOF (also created by the same aperture as creates diffraction) is certainly an overwhelming part of that

I double dare you, do actually look at the f/40 images I pointed you at. Neither is much of an image, but try to overtly tell me that the f/11 image is specifically better than the f/40 image.

Sure, it is a special case, and sure, f/40 would be a terrible choice for most routine images of the kids, but there are times it is extremely useful and helpful. There are even more times for f/22 and f/32. And sometimes we may choose f/1.8 (frankly, I rarely do). It all depends.

But to advise anyone that they should never venture past f/11 or f/16 because of diffraction is really terrible advice.

 

[J-see]

Aw come on J-see, I doubt you really are that obtuse. The issue is about conditions for creating a better image, and diffraction generally detracts (true of the step from f/8 to f/11 too). However DOF (also created by the same aperture as creates diffraction) is certainly an overwhelming part of that

I double dare you, do actually look at the f/40 images I pointed you at. Neither is much of an image, but try to overtly tell me that the f/11 image is specifically better than the f/40 image.

Sure, it is a special case, and sure, f/40 would be a terrible choice for most routine images of the kids, but there are times it is extremely useful and helpful. There are even more times for f/22 and f/32. And sometimes we may choose f/1.8 (frankly, I rarely do). It all depends.

But to advise anyone that they should never venture past f/11 or f/16 because of diffraction is really terrible advice.

To be honest Wayne, I looked at the detail shots on your page two and even while you declare f/22 the winner in many cases, I clearly see that the part in focus is noticeable softer which is what it is about here. It's however possible something is wrong with my vision but I'll let others be the judge of that.

I said that the smaller the sensor pixels, the more important fast lenses become if you want to have some options left. Technically they could go as small as 1µ with those BSI sensors but they'll become severely diffraction limited. Resolution and detail would suffer when using too slow lenses. DoF has nothing to do with this; it's besides the point.

If you want to crop small or print large with such sensors, you have to be able to open up wide.